substantive stochastic tests

[sbenthall]

HARK does not work unless it can replicate certain substantive economic results.

Currently, some of these tests (in deterministic cases) are intermingled with unit tests and run automatically.

For tests based on stochastic results (for example, the mean across many simulations), it is not appropriate to put these tests in the same place as the unit tests. That's because unit tests should not stochastically fail.

However, it would be good to have an automated test suite for substantive results that includes these stochastic tests.
These might be run with a separate command from the unit tests.
It would also be better to separate these so that it's easier for Economists to review them. (They might include more in-depth description of what they represent, references to the literature, etc.)

[sbenthall]

One example would be a minimized version of the cstwMPC results, with an infinite horizon agent instead of a lifecycle model.